While tight glucose control has been shown to improve the outcomes of some critical care patients, much controversy regarding its overall benefit persists;in part due an unacceptable incidence of hypoglycemia, or low blood sugar. A decision support system for glucose control in critical care, much like an artificial pancreas, is comprised of three essential components: (1) a glucose measuring device, (2) an algorithm that interprets this measurement and recommends a treatment strategy, and (3) a delivery device that implements this strategy, delivering insulin, glucose, or some other agent (e.g., glucagon) to a patient. This proposal will use systems engineering tools to provide a robust answer to the following questions: given the characteristics of a minimally invasive glucose measuring device, what is the tightest glucose control achievable while avoiding hypoglycemia, and what is the strategy to achieve this control? We propose to use a very large multi-center dataset of critically ill patients receiving insulin, aiming to (1) calibrate and validate a mathematical model of glucose and insulin dynamics and (2) characterize between-patient variations as embodied in model parameters. Such a model will then be used to (3) design and deliver a patient-tailored decision support system, in the form of a portable interface that would forewarn clinical practitioners of potential hypoglycemic episodes and recommend insulin or dextrose dose administration. The ultimate goal of this proposal is to put all necessary tools in place for a randomized clinical trial of tight glucose control in critically ill patients, while completely avoiding episodes of hypoglycemia. It is expected that a successful completion of this proposal will have high translational impact and contribute to systems engineering science, specifically in the tailoring of sophisticated algorithms to patient- specific needs.
Critically ill surgical and medical patients demonstrate better survival with tight glucose control, but tight glucose control in clinical trial populations has often been achieved with an unacceptable rate of hypoglycemic (low blood sugar) episodes requiring further treatment. The research program proposed will develop an interactive model-based Decision Support System that would forewarn clinical practitioners of potential hypoglycemic episodes and recommend insulin or dextrose dose administration. The ultimate goal of this proposal is to put all necessary tools in place for a randomized clinical trial of tight glucose control in critically ill patients, while completely avoiding episodes of hypoglycemia via our decision support system.
